Le lundi 16 juin 2014 à 14h, amphithéâtre P, bât. 12D -campus de Beaulieu Université de Rennes 1, Monsieur Juan-David OSPINA ARANGO soutient son doctorat en cotutelle, mention Traitement du Signal et Télécommunications, intitulé :"Predictive models for side effects following prostate cancer radiotherapy", devant le Jury composé de : M. Gilles CRÉHANGE (Rapporteur; PU-PH, Centre Georges François Leclerc, Dijon), M. Jean-Yves TOURNERET (Rapporteur; PU, INP-ENSEEIHT, Toulouse), M. Jean-Léon LAGRANGE (Examinateur; PU-PH, AP-AH Hôpital Henri Mordor, Créteil), M. François KAUFFMANN (Examinateur; MCU Université de Caen - Basse Normandie, Caen), M. Raúl Alberto PÉREZ AGÁMEZ (Examinateur, Professeur, Universidad Nacional de Colombia, Medellin, M. Oscar ACOSTA (Co-directeur de thèse, MCU, LTSI), M. Juan Carlos CORREA MORALES (Co-directeur de thèse; Professeur, Universidad Nacional de Colombia, Medellin) et M. Renaud de CREVOISIER (Directeur de thèse PU-PH).
Résumé : External beam radiotherapy (EBRT) is one of the cornerstones of prostate cancer treatment. The objectives of radiotherapy are, firstly, to deliver a high dose of radiation to the tumor (prostate and seminal vesicles) in order to achieve a maximal local control and, secondly, to spare the neighboring organs (mainly the rectum and the bladder) to avoid normal tissue complications. Both objectives are, however, in conflict and a compromise is needed to achieve an effective treatment and maintain a good quality of life after prostate EBRT. Normal tissue complication probability (NTCP) models are then needed to assess the feasibility of the treatment and inform the patient about the risk of side effects, to derive dose-volume constraints and to compare different treatments. Proposed in the 1970’s, the classic NTCP models can be modified to be adapted to both biological paradigm and treatment technique changes. The NTCP models have evolved from empirical models with parameters estimated via consensus to machine-learning methods trained on real data. In the context of EBRT, the objectives of this thesis were to find predictors of bladder and rectal complications following treatment; to develop new NTCP models that allow for the integration of both dosimetric and patient parameters; to compare the predictive capabilities of these new models to the classic NTCP models and to develop new methodologies to identify dose patterns correlated to normal complications following EBRT for prostate cancer treatment. A large cohort of patient treated by conformal EBRT for prostate cancer under several prospective French clinical trials was used for the study. In a first step, a traditional statistical regression approach was used to find predictors of bladder and rectal complications following EBRT. Using Kaplan-Meier nonparametric estimation, the incidence of the main genitourinary and gastrointestinal symptoms have been described. With another classical approach, namely logistic regression, some predictors of genitourinary and gastrointestinal complications were identified. The logistic regression models were then graphically represented to obtain nomograms, a graphical tool that enables clinicians to rapidly assess the complication risks associated with a treatment and to inform patients. This information can be used by patients and clinicians to select a treatment among several options (e.g. EBRT or radical prostatectomy). In a second step, the difficulty of including both dosimetric and patient parameters in classical NTCP models was identified. Although this can be done by stratifying a population and then fitting a different model at each stratum, this stratification may lead to a loss of statistical power as not all strata met the required number of patients to fit the models. Another strategy is adding new parameters to existing models, but this implies making biological assumptions which are difficult to confirm with treatment parameters and patient outcome data only. We thus proposed the use of random forest, a machine-learning technique, to predict the risk of complications following EBRT for prostate cancer. Random forest show a similar accuracy as the most accurate state-of-the-art machine-learning methods but without overfitting. Our random forest NTCP (RF-NTCP) model, which includes both clinical and patient parameters, was compared to traditional Lyman-Kutcher-Burman (LKB) NTCP and logistic regression models. The superiority of the RF-NTCP, assessed by the area under the curve (AUC) of the receiving operative characteristic (ROC) curve, was established. In a third step, the 3D dose distribution was studied. A 2D population value decomposition (PVD) technique was extended to a tensorial framework to be applied on 3D volume image analysis. Using this tensorial PVD, a population analysis was carried out to find a pattern of dose possibly correlated to a normal tissue complication following EBRT. Also in the context of 3D image population analysis, a spatio-temporal nonparametric mixed-effects model was developed. This model was applied to find an anatomical region where the dose could be correlated to a normal tissue complication following EBRT. In conclusion, the main contribution of our work is the development of new predictive models of rectal and bladder toxicities following EBRT and the demonstration that these models are strong competitors of the classic NTCP models.